Anchored Matching Adjusted Indirect Treatment Comparison of Quizartinib Vs Midostaurin in Newly Diagnosed Patients with FLT3-ITD-Positive Acute Myeloid Leukaemia

Introduction Quizartinib and midostaurin are FLT3 inhibitors approved for newly diagnosed (ND) fit FLT3-ITD+ AML based on QuANTUM-First (Q-F) and RATIFY trials. There is currently no randomized controlled trial (RCT) directly comparing the efficacy of quizartinib vs midostaurin in these patients. Indirect treatment comparisons (ITC) have been used to compare the relative effects of treatments in the absence of RCTs directly comparing the interventions. However, matching adjusted indirect comparisons (MAIC) are preferred over unadjusted ITC methods in the presence of significant heterogeneity between the trials. We performed an anchored MAIC to estimate the relative efficacy of quizartinib vs midostaurin for patients with ND FLT3-ITD+ AML. Methods Anchored MAIC was preferred over unanchored due to both trials having a comparable placebo arm. Individual patient-level data was available for the Q-F trial, while aggregate data for complete remission (CR), relapse after CR (CIR) overall survival (OS), and baseline characteristics were available for the FLT3-ITD-+ subgroup of RATIFY trial. Treatment effect modifiers (TEMs) were identified based on interaction analyses and validated by clinical experts. Propensity score weighting method was used to create a matched cohort of FLT3-ITD+ Q-F patients to compare vs. RATIFY patients. OS was estimated using Cox proportional hazards models, while CR was compared using odds ratios (OR). Estimation for CIR was done using a Fine-Gray competing risk model as death was a competing event to relapse. Significance was assessed via 95% confidence intervals. Naïve (unadjusted) comparison and MAIC results are reported. Results Compared to the RATIFY, the Q-F patients had a higher median age (56.0 vs 47.1 yrs), did not include FLT3-TKD patients, had lower platelet counts (22 vs 50x103), and lower absolute neutrophil count/mm3 (0.1 vs 2.2) making comparison between the two studies difficult. The Q-F ITT population was matched to the RATIFY subgroup using four TEMs - platelet count at baseline, sex, age at baseline, and NPM1 mutation status, resulting in an effective sample size (ESS) of 54% of original sample size. The OS hazard ratio (HR) numerically favoured quizartinib (Naïve 0.87; 95% CI: 0.56-1.34; MAIC 0.82; 95% CI: 0.48-1.39), while the CR OR was statistically non-significant with wide confidence intervals (Naïve 0.83; 95% CI: 0.44-1.56; MAIC 0.92; 95% CI: 0.42-1.97). The CIR HR favoured quizartinib (Naïve 0.61; 95% CI: 0.31-1.19; MAIC 0.42; 95% CI: 0.20-0.91) which was statistically significant. Conclusion After accounting for differences in the trial populations, although statistically significant CR and OS differences were not observed, the MAIC revealed that quizartinib generated a significantly reduced risk of relapse which suggests a deeper and longer duration of CR. The results from the MAIC should be interpreted with care due to remaining residual heterogeneity stemming from potential TEMs not matched for, due to limitations of reported data from RATIFY.